As part of the “Research and Innovation” joint booth organized by the Ministry of Science and Health (MWG) Rhineland-Palatinate, i3mainz presented a project from its AI lab in collaboration with TOPML ( “Trading Off Non-functional Properties of Machine Learning”) at Johannes Gutenberg University (JGU) at the Hannover Messe 2026 from April 20 to 24.
The joint booth of the State of Rhineland-Palatinate, Photo: Julian Stockemer, CC BY-SA 4.0
Digital Crime Scene Analysis with AI and Virtual Reality
Julian Stockemer represented i3mainz at the event. He presented his concept for language-based interaction and spatial contextualization in digital crime scene analysis. The challenge lies in the fact that modern crime scene reconstructions generate large amounts of heterogeneous and sensitive data that can often only be linked to a limited extent and must not be transferred to external AI systems. Stockemer presented his prototypical approach to a multimodal VR application for crime scene analysis. By combining spatial and semantic information, a digital twin of the crime scene is created, enabling natural interactions through speech and gaze direction. At the same time, the approach aims to create transparency in the analysis process and support the efficient evaluation of complex data.
AI Language Models and Gender Bias
With their exhibit “LLMs Can Now Listen: And Produce Stereotypes Based on Voice,” Xenia Heilmann and Kirsten Köbschall from JGU presented current research findings on audio-capable language models. The researchers were able to demonstrate that models derive characteristics—such as the speaker’s gender—from vocal features and generate stereotypical responses based on them. The results highlight a trade-off between efficiency and fairness: The more accurately a model identifies gender, the more strongly gender-specific biases can manifest in the generated responses.
TOPML and the AI Lab
TOPML investigates key properties of machine learning systems, such as transparency, fairness, data protection, and resource efficiency. The project’s goal is to highlight interactions and trade-offs between these properties and to develop solutions that take ethical and legal requirements into account.
Through the AI Lab, i3mainz aims to transfer research findings into industrial practice. With their presentation at the 2026 Hannover Messe, Johannes Gutenberg University Mainz and Mainz University of Applied Sciences demonstrated the importance of research into responsible and application-oriented artificial intelligence.
